The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends that users install the 'Agent Reach' tool from an untrusted GitHub repository (https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md). Although this is a documentation reference, it encourages the adoption of third-party software from a source that is not included in the trusted vendor list.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection due to its automated integration of web search results into the agent's context.
Ingestion points: Untrusted data enters the agent context via the output of the exa.web_search_exa tool, as described in the search fallback strategy in SKILL.md (Step 6).
Boundary markers: The skill does not implement specific delimiters or instructions to the agent to ignore potentially malicious commands embedded in the retrieved search content.
Capability inventory: Across its documentation and logic, the skill demonstrates capabilities for network tool invocation (Exa search) and reading local knowledge base files.
Sanitization: There is no evidence of filtering or sanitization being applied to the external search data before it is processed by the model to generate a response.
[SAFE]: The skill utilizes Exa.ai (https://mcp.exa.ai/mcp), which is a well-known service for AI search applications, for its web search fallback functionality.
[SAFE]: All identified Huawei Cloud resources, such as console.huaweicloud.com and help center links, are legitimate vendor assets and do not pose a security risk.

huawei-cloud-icp-process-guidance